Skip to main content

Configure generation

At least three characters: only letters, numbers, spaces and hyphens allowed.

Maximum number of keywords to analyze in the DBSCAN clustering process.

Come utilizzare Content Cluster Builder AI

Insert a keyword seed

Write the starting topic (e.g. "EU AI Act compliance"): at least 3 characters, only letters, numbers, spaces, and apostrophes.

Set depth keyword and language

Set the depth slider to decide how many keywords to analyze in the clustering DBSCAN, and choose your desired output language.

Generate semantic clusters

Click "Generate Cluster": the backend queries the RediSearch HNSW vector index of the portfolio and uses Ollama on-prem to classify the intent (informative, transactional, or navigational) and propose a markdown article outline.

Explore, validate and export results

Open each cluster to view keywords, outlines and suggestions for internal linking to existing articles. Export all in JSON to integrate into your editorial workflow editor.

Suggerimenti

  • Use a low depth (10-20) to quickly validate an idea of content, and a high depth (70-100) only when you want to plan an entire pillar content with multiple article clusters.
  • Follow all internal linking suggestions: linking semantically related articles strengthens the site's topical authority in search engines' eyes.

Domande frequenti

How are clusters of keywords generated?

Backend expands the seed keyword, calculates vectorial embeddings and applies DBSCAN clustering to group semantically similar terms. Each cluster is then enriched with Ollama (llama3.1:8b on-prem) with intent classification and article outline classification.

What do information, transaction, and navigation labels mean?

Informational - "I want to know" type research (guides, explanations). Transactional - Research related to an action or conversion (comparisons, purchases, subscriptions). Navigational - Search for a known brand or tool.

Are AI-generated results reliable for an SEO strategy?

This is a useful starting point for idea generation, but must always be validated with professional SEO tools (Ahrefs, SEMrush) before making definitive editorial decisions, as indicated in the visible disclaimer below results.

Are my data sent to external cloud services?

No, the entire processing (embedding, clustering and outline generation) happens on-prem on our VPS via Ollama; no data is sent to OpenAI, Anthropic or other third-party cloud providers.

What to do if the depth keyword returns few clusters?

If the RediSearch vector index finds few semantically close keywords to the inserted seed, DBSCAN clustering may produce few or very small clusters. Try increasing depth (slider towards 70-100) or using a broader, less niche keyword seed.